TL;DR
This study evaluated the cost-effectiveness of a data-driven, personalized antibiotic dosing system (AutoKinetics) compared to standard care in critically ill sepsis patients. While the system showed some promise in improving pharmacokinetic target attainment, it did not prove to be a cost-effective intervention, suggesting current models may not justify widespread implementation.
What Problem Does This Paper Address?
Sepsis and septic shock are life-threatening conditions affecting millions globally, with high mortality rates and significant economic burdens. Effective treatment relies on timely and appropriate antibiotic administration. However, critically ill patients exhibit altered pharmacokinetic profiles, making standard dosing regimens suboptimal. This can lead to toxicity from overdosing or treatment failure and antimicrobial resistance from underdosing. The AutoKinetics system was developed to address this by providing real-time, model-informed, personalized antibiotic dosing recommendations at the bedside. This study aimed to determine if this advanced approach is economically viable compared to usual care.
Methods
The researchers conducted a cost-effectiveness analysis alongside a two-center, non-blinded, randomized controlled trial (AutoKinetics trial). The economic evaluation adopted a societal perspective, considering all costs incurred by patients and healthcare systems. Key outcomes measured were:
- Quality-Adjusted Life Years (QALYs): Assessed using the EQ-5D-5L questionnaire at ICU discharge and 6 months post-discharge, with linear interpolation for missing time points.
- Mortality: Tracked at 6 months post-ICU discharge.
- Pharmacokinetic Target Attainment (TA): Measured for the primary antibiotic within the first 24 hours of treatment, defined as achieving 75% of the dosing target.
Costs included ICU and general ward stays, healthcare utilization post-discharge, and productivity losses (presenteeism and absenteeism). Statistical analysis involved multiple imputation for missing data and bootstrapping to estimate uncertainty. Incremental Cost-Effectiveness Ratios (ICERs) were calculated, and cost-effectiveness acceptability curves (CEACs) were generated to assess the probability of cost-effectiveness at various willingness-to-pay thresholds.
Datasets
The study utilized data collected from patients enrolled in the AutoKinetics randomized controlled trial conducted at two tertiary referral centers in Amsterdam, the Netherlands. No external datasets were used for training or evaluation.
Key Insights and Ideas
- Personalized Dosing is More Expensive: The AutoKinetics group incurred higher overall healthcare costs, primarily driven by longer ICU stays, despite some mitigation from lower productivity losses.
- Limited Statistically Significant Improvements: While the AutoKinetics group showed trends towards lower mortality and more QALYs, only the improvement in pharmacokinetic target attainment (TA) was statistically significant.
- High ICERs Suggest Poor Value: The cost to gain one QALY or prevent one death with AutoKinetics was substantial, exceeding typical willingness-to-pay thresholds.
- TA is a Niche Outcome: While TA improved significantly, the lack of a consensus on a willingness-to-pay threshold for this specific outcome makes its cost-effectiveness difficult to interpret.
- Study Limitations Impact Interpretation: The trial's early termination due to COVID-19, potential underestimation of QALYs due to interpolation, and assumptions about productivity losses in non-survivors introduce uncertainty and may bias results against the intervention.
Key Results
- Costs: AutoKinetics group: €42,684 vs. Control group: €39,475.
- Mortality: AutoKinetics group: 42% vs. Control group: 49%.
- QALYs: AutoKinetics group: 0.184 vs. Control group: 0.153.
- Pharmacokinetic Target Attainment (TA): AutoKinetics group: 69% vs. Control group: 48%.
- ICERs (Societal Perspective, Adjusted Analysis):
- QALY: €123,493 per QALY gained.
- Mortality: €55,576 per death prevented.
- TA: €18,129 per patient achieving target attainment.
- Probability of Cost-Effectiveness: For QALYs and mortality, the probability of cost-effectiveness was generally below 60% for most acceptable willingness-to-pay thresholds.
Limitations and Future Work
The authors acknowledge several limitations:
- Underpowered Trial: The trial was stopped early, limiting its power to detect statistically significant differences in costs and outcomes, particularly for QALYs and mortality.
- QALY Estimation: Linear interpolation of utility scores between ICU discharge and 6 months post-discharge may underestimate actual QALYs, especially for the AutoKinetics group with higher survival.
- Productivity Loss Assumption: Assuming zero productivity costs for deceased patients might disproportionately affect the AutoKinetics group due to its higher survival rate, though the impact is likely mitigated by the older patient population.
- Future Software Costs Unknown: The cost of the AutoKinetics software itself was not factored in, as it was still under development.
- High Missing Data: A significant proportion of data for QALYs and costs was missing, requiring imputation.
Future research should focus on larger, adequately powered trials to confirm findings. Measuring health-related quality of life at more time points and reducing missing data are crucial. Further development of pharmacokinetic models and a clearer understanding of the value of pharmacokinetic target attainment are also recommended.
Summary updated